Researchers from Anyscale and NVIDIA demonstrated a method to scale robot policy evaluation by disaggregating GPU-heavy simulation and policy inference workloads using Ray and Isaac Lab. The approach …
Microsoft announced a broad set of enhancements to Azure Kubernetes Service at Build 2026, including bare metal support, fleet management, and AI infrastructure features, aiming to make Kubernetes a f…
Google Cloud and Anyscale announced optimizations for Ray Serve LLM on Google Kubernetes Engine (GKE) that deliver up to 5x higher throughput and 8x lower latency for large language model inference. T…
Oxlo.ai offers request-based pricing for AI inference, charging a flat fee per API call regardless of prompt length, contrasting with token-based models used by providers like Together AI and Firework…
A developer compared LLM inference APIs on cost, performance, and integration, noting that most providers use token-based pricing which can cause unpredictable costs for long-context or agentic worklo…
Engineers achieved up to 67% cost savings and 2.7x better goodput by using Prefill-Decode disaggregation with Ray and vLLM on AMD MI325X GPUs, separating prefill and decode phases onto dedicated hardw…
Ray Serve LLM and vLLM on AMD MI325X achieve up to 67% cost savings by disaggregating prefill and decode phases in LLM serving, separating them onto dedicated GPUs to eliminate interference and improv…
Anyscale released agent skills for debugging Ray workloads, including /anyscale-platform-fix and /anyscale-platform-inspect, which automate troubleshooting of failing pipelines. A user demonstrated fi…
Adyen trained a Transaction Foundation Model on 51 trillion tokens using Ray, while Xoople, Criteo, and BMW shared their own scaling AI stories at Anyscale's Ray Day London event. The event highlighte…
At Anyscale's Ray Day: NYC event, Torc Robotics reported achieving 90% GPU utilization by consolidating its fragmented multimodal AI stack onto Ray, up from 30-40%. Discord detailed its ML platform ev…
A benchmark by Alluxio and Anyscale shows that using Alluxio as a distributed NVMe cache for Ray Data reduces cross-region training data read times from 4,241 seconds to 208 seconds, a 20x speedup, by…
Anyscale on Azure entered public preview, allowing Azure customers to provision the AI compute platform powered by Ray inside their own tenancy. Co-engineered with Microsoft, the integration inherits …
Trajectory, in collaboration with UC Berkeley Sky Lab and Anyscale, released a concurrent multi-LoRA training stack for continual learning that achieves a 2.81× end-to-end experiment-throughput gain o…
Trajectory, in collaboration with UC Berkeley Sky Lab and Anyscale, built a concurrent, multi-LoRA training platform for continual learning workloads that achieved a 2.81× end-to-end experiment-throug…
GPU utilization rates on major clouds average just 5 percent, meaning most paid GPU hours produce no useful work, while egress fees of up to $0.12 per GB and data gravity lock-in add hidden costs that…
Anyscale released Agent Skills, a token-efficient tool for building Ray pipelines, and introduced a new maturity model for ML operations. The skills act as first responders across build, deploy, and o…